Dynamic temporal reinforcement learning and policy-enhanced LSTM for hotel booking cancellation prediction

Area of Science:

• Computer Science
• Artificial Intelligence
• Data Science

Background:

• The global tourism industry's rapid expansion necessitates efficient hotel booking management.
• Traditional cancellation prediction models fail to account for dynamic temporal factors like seasonality and events.
• Existing methods struggle with the dynamic nature of real-world cancellation behaviors.

Purpose of the Study:

• To develop a novel deep reinforcement learning framework for accurate hotel booking cancellation prediction.
• To address the limitations of static models in capturing temporal dynamics and market fluctuations.
• To enhance hotel service efficiency and operational management through improved prediction accuracy.

Main Methods:

• Implementation of a novel framework combining dynamic temporal reinforcement learning with policy-enhanced Long Short-Term Memory (LSTM).
• Leveraging multi-source information to capture complex temporal dynamics and improve prediction stability.
• Utilizing deep reinforcement learning techniques for adaptive and accurate forecasting of cancellation trends.

Main Results:

• The proposed model achieved over 95.9% prediction accuracy, significantly outperforming traditional methods.
• Demonstrated high model stability (0.98), an F1 Score approaching 1, and a mutual information score of approximately 0.93.
• Validated effectiveness and generalization across diverse data sources, confirming robustness.

Conclusions:

• Deep reinforcement learning offers an innovative and efficient solution for managing hotel booking cancellations.
• The proposed model effectively handles complex prediction tasks with dynamic temporal influences.
• This approach enhances the potential of AI in optimizing operational efficiency within the hospitality sector.